"Unique Visitors" doesn't have much to do with people
I was taking my morning coffee, reading Brian Clifton post about "Should you focus on website visitors as individuals?", his views of a recent e-Consultancy interview of Coremetrics Chief Strategy Officer, John Squire. I commented on Brian's blog, but as is often the case, my thoughts turned into something too long for a comment. I strive to participate in conversations I know won't end up with the now classic "let's agree to disagree".
(emphasis mine) I salute Coremetrics (and others) objectives to track "unique people" instead of "unique cookies" in the hope to get more precise and detailed information. But this is a lost cause... the inherent technologies of the web don't allow that, unless you can authenticate every single visitor coming through all channels... As soon as there is "multichannel" and "web analytics" in the same phrase I get suspicious... "multichannel" when everything is online is something, but how is data from call center, back-end core systems and other touch points integrated to give a "single view of the customer"? We're heading in the realm of business intelligence (BI) and customer relationship management (CRM), and despite the marketing fluff, I don't see any of the current web analytics vendor really playing in this space. In the above statement, all of the mentioned "channels" are "online", which, I agree, gives you a very good view to start with, but a view where your sight is concentrated and everything else in the periphery is blurred.
I like Brian argument that aggregating & sampling is the way to go. After all, web analytics is closer to statistics: with a population, segments, sampling/aggregating... and a margin of error, confidence and significance... then it is to core systems where each piece of data should be 100% accurate and traceable.
From time to time, access to the raw data, or the lowest level possible, is justifiable and necessary. The latest Forrester Wave: Web Analytics Q3 2009 reveals that 49% of surveyed web analytics clients said "accuracy of information" is one of their three most important factors when selecting a vendor. This comes ahead of reporting options, and far ahead of TCO! Yet, despite vendors secret formula to calculate accurate visits & visitors, research I conducted using WASP shows the number one source of distrust and inaccurate data comes from inappropriate and incorrect instrumentation.
Tracking individual customers, really?
Coremetrics is differentiating its technology by focusing on individual customer data in a multichannel environment. They look at “complete historical online behavior and brand interactions on websites, across multiple ad networks and via email, video, affiliate sites, social media, and more”.
(emphasis mine) I salute Coremetrics (and others) objectives to track "unique people" instead of "unique cookies" in the hope to get more precise and detailed information. But this is a lost cause... the inherent technologies of the web don't allow that, unless you can authenticate every single visitor coming through all channels... As soon as there is "multichannel" and "web analytics" in the same phrase I get suspicious... "multichannel" when everything is online is something, but how is data from call center, back-end core systems and other touch points integrated to give a "single view of the customer"? We're heading in the realm of business intelligence (BI) and customer relationship management (CRM), and despite the marketing fluff, I don't see any of the current web analytics vendor really playing in this space. In the above statement, all of the mentioned "channels" are "online", which, I agree, gives you a very good view to start with, but a view where your sight is concentrated and everything else in the periphery is blurred.I like Brian argument that aggregating & sampling is the way to go. After all, web analytics is closer to statistics: with a population, segments, sampling/aggregating... and a margin of error, confidence and significance... then it is to core systems where each piece of data should be 100% accurate and traceable.
Trust me, I'm accurate (and honest)
One of the thing playing against Google Analytics is the fact that experienced analysts (those who have a statistical background, come from the offline world or have lots of experience in a real multichannel, multi-system environment) want to audit, validate and diagnose the data they are relying on. The inability to audit the data and have a peak at the methodologies employed by Google to provide the data is a big showstopper to some organizations. It's basically asking for a blind trust where your provider (AdWords) is also giving you your results (analytics conversions)... or your employer is also managing your bank account... For that mater, your also have to trust other vendors because you can't have access to the algorithms they use for aggregating or sampling data, but they do not play a dominant role in your online marketing (PPC) spends.From time to time, access to the raw data, or the lowest level possible, is justifiable and necessary. The latest Forrester Wave: Web Analytics Q3 2009 reveals that 49% of surveyed web analytics clients said "accuracy of information" is one of their three most important factors when selecting a vendor. This comes ahead of reporting options, and far ahead of TCO! Yet, despite vendors secret formula to calculate accurate visits & visitors, research I conducted using WASP shows the number one source of distrust and inaccurate data comes from inappropriate and incorrect instrumentation.

3 comments:
Great points Stephane
In particular, I find time and time again, accuracy issues come down to how well the implementation has been done, not the tool itself.
Lets face it, tracking visitors using cookies and javascript has been around 10+ years and is not rocket science. In the early days, I even wrote a simple tracking script myself! So in terms of data collection, all the major vendors are on the same footing.
In terms of "details", I have truthfully not met a client yet that says "hey - there is not enough detail in these GA reports, give me more". It is very much the other way round.
In addition, there is so much low hanging fruit when it comes to improving a website, that zooming in too far is counter productive - see: http://www.advanced-web-metrics.com/blog/2009/02/22/improving-the-web-with-web-analytics/
That said, it would be great to hear on what additional details you think GA is missing. Another blog post?
Brian: I agree, making the best use of what GA already offers is enough in most cases. I'm a proponent of taking enlightened decision with what we have rather than becoming paralyzed by the specter of mistakes and perfection.
My POV about what's missing in GA: custom attributes at the page and visit/visitor level. That's about it. Check out my 3 points in Playing with Google Analytics Pivot Tables
Thanks for your comment! :)
Hi Stephane,
I completely agree, and funnily enough I too just wrote a post on data accuracy in web analytics
Also, as a good example - I recently took a free trial of some Bose headphones (sent them back though, rip-off). Whilst deciding whether to do this I visited the site several times: once from work, once from my home PC, once from my home netbook, and another time from my iPhone. How could Coremetrics think anything else than I am four different people??
Jonny Longden
Actionable Analytics Blog
Post a Comment